Transfer from: Prospective Industry Research Institute

major listed companies in the industry: Baidu group (09888), Tencent holdings (00700), wave information (000977.SZ), zhongke shuguang (603019.SH), huasheng tiancheng (600410.SH), qingyun technology (688316.SH), etc.

the core data of this article: the current situation of AI computing power development; AI computing power expenditure; The scale of computing power; Intelligent computing power; Regional competition pattern

Global AI Computing Power Development Status

At present, the current situation of the global AI computing industry has four characteristics: the proportion of global AI computing expenditure has increased, the scale of global computing power has grown rapidly, the global computing power segment has changed from basic computing power to intelligent computing power, and the growth rate of the latter has accelerated; the demand for global computing power has increased significantly.

The rise of new demands and new businesses, such as artificial intelligence applications and large model training, represented by AIGC, has profoundly affected the changes in global economic and social development, promoting the rapid growth of computing power, diversified innovation in computing technology and accelerated restructuring of the industrial landscape. The productivity role of computing power in boosting the development of the global digital economy has become more prominent, and it has become a new focus that cannot be ignored in the strategic competition of various countries.

Multi-technology collaborative upgrade promotes the continuous development of advanced computing. On the one hand, the evolution of computing technology is accelerating, and heterogeneous computing has become the mainstream architecture of intelligent computing cycle high computing power. Against the backdrop of Moore's Law slowing down and subversive technology not yet mature, multi-application innovation-driven computing, represented by the AI Big Model, accelerates into a new cycle of intelligent computing, further driving the restructuring and reshaping of the computing industry landscape.

In the era of intelligent computing, AI servers and on-board computing platforms equipped with various computing acceleration chips will become the main sources of computing power. On the other hand, the innovation of advanced computing system is active, and the innovation mode and focus are transformed, showing the characteristics of soft and hard integration and system architecture innovation. Technological innovation continues to cover the basic process, hardware, software, the whole machine at different levels including 4nm and 3nm process upgrades, interconnection continues to high-speed, cross-platform evolution, soft and hard coupling to accelerate intelligent computing into the E-level era.

global AI computing power expenditure accounts for nearly 25%

according to the data of the 2022-2023 global computing power index evaluation report, from the perspective of 15 major countries in the world, the proportion of AI computing power expenditure in total computing power expenditure increased from 9% in 2016 to 12% in 2021, with preliminary accounting accounting for 18% in 2022 and expected to reach 25% by 2025.

Changes in global computing power scale: rapid growth

The scale of global computing power has maintained rapid and stable growth. In the context of the digital economy era characterized by the perception of all things, the interconnection of all things, and the intelligence of all things, the total amount of global data and the scale of computing power continue to show rapid growth. According to IC data, total global data production in 2022 is 81ZB.

According to the White Paper on China's Computing Power Development Index released by the China Academy of Information and Communications, the global computing power scale has maintained a high and stable growth trend in recent years, with the total computing power of global computing equipment reaching 615 EFlops in 2021, a growth rate of 44%. In 2022, the total computing power of global computing equipment will reach 906EFlops, with a growth rate of 47%.

Global Computing Power Segments: Intelligent Computing Power Develops Vigorously

The global computing power segment is divided into basic computing power, intelligent computing power and supercomputing power. In 2021, the basic computing force scale (FP32) is 369 EFlops, the intelligent computing force scale (converted to FP32) is 232 EFlops, and the supercomputing force scale (converted to FP32) is 14 EFlops. In 2022, the basic computing force scale (FP323) is 440EFlops, the intelligent computing force scale (converted to FP32) is 451EFlops, and the supercomputing force scale (converted to FP32) is 16EFlps. In terms of market segments, the development of smart computing is accelerating, with a growth rate of 94.4 percent in 2022, and the scale of computing power exceeds the basic computing power.

Note: The inner circle is 2021 data and the outer circle is 2022 data.

① The scale of basic computing power is estimated according to the total amount of server computing power in the world in the past six years. Global basic computing power = ∑ last six years (annual server shipment size * average server computing power for the year).

② FP32 is a single-precision floating-point number, FP16 is a half-precision floating-point number, and FP64 is a double-precision floating-point number.

③ The scale of intelligent computing power is estimated according to the total amount of AI server computing power in the world in the past 6 years. Global intelligent computing power = ∑ last six years (annual AI server shipment size * average AI server computing power for that year).

④ The scale of supercomputing power is mainly based on the global supercomputer TOP500 data, and is estimated with reference to the relevant data of supercomputing manufacturers.

China and the United States are the world's major computing powers

In terms of the distribution of global computing power, according to the China Institute of Information and Communications, the shares of the United States, China, Europe and Japan in the global computing power scale will be 34%, 33%, 14% and 5%, respectively, in 2021. Among them, the global basic computing competition to the United States and China as the first echelon, and the gap between China and the United States is narrowing.

In 2022, global computing competition will intensify and globalization will face challenges. In terms of computing power scale, the share of the United States, China, Europe and Japan in the global computing power scale is 34%, 33%, 17% and 4%, respectively, as measured by the China Institute of Information and Communications, while the share of the United States and China is the same as in 2021.

For more research and analysis of this industry, please refer to the "China AI Computing Power (Artificial Intelligence Computing Power) Industry Development Prospect and Investment Strategic Planning Analysis Report" by the Prospective Industry Research Institute.

at the same time, the prospective industrial research institute also provides solutions such as industrial big data, industrial research report, industrial planning, park planning, industrial investment promotion, industrial map, intelligent investment promotion system, industry status certificate, IPO consultation/feasibility study of investment raising, IPO working paper consultation, etc. Citing the contents of this article in any public disclosure such as the prospectus and the company's annual report requires formal authorization from the Prospective Industry Research Institute.

more in-depth industry analysis can be found in [prospective economist APP], and you can also communicate and interact with 500 + economists/senior industry researchers.